The Immunology Team, which is now part of the Lymphocyte Biology Section, Laboratory of Systems Biology, reported previously on a top-down analysis of the immune and non-immune tissue response of mice to various strains of influenza virus. Highly standardized preparations of both mildly pathogenic (Tx91) and highly pathogenic (PR8) viruses were used at varying infectious doses in a single inbred strain of mouse and several hundred highly qualified microarray transcriptional analyses conducted with RNA isolated from infected lung tissues of these mice at varying time points post-inoculation. Modular gene set analysis uncovered marked differences in transcriptional responses with the PR8 vs. Tx91 strains of virus that can be assigned to specific biological processes. Principal component analysis (PCA) revealed that the transcriptional responses to infection results in cell-type specific changes that largely reflect cell recruitment into the lungs and also context (infection-type) specific changes within the cell-specific gene sets. Furthermore, gene sets related to inflammatory responses are the major component associated with lethality, whereas anti-viral gene sets are similar with both the low and high pathogenicity viruses. A positive feedback pathway involving virus-induced chemokine production facilitates recruitment of myeloid cells to the lungs. Together, these data suggested that uninterrupted amplification of myeloid cell recruitment and inflammatory cytokine production plays a key role in pathogenic infections. In support of this model, attenuating but not eliminating myeloid cell recruitment using depleting antibodies rescues mice from early lethality of PR8 infection. Thus, this study uncovered a core feedback circuit involving innate inflammation that drives early lethality in influenza infection and provides new targets for intervention in this disease. We are now extending this work in an effort to develop a robust computational framework for utilizing peripheral blood transcriptomic data to define tissue-specific pathology. To date, this has proven very difficult, in large measure because the multiple testing correction needed when using whole genome transcriptional data and no starting hypothesis can hide small but significant signals within the blood signature. We tested the hypothesis that a tissue-based definition of a lethal signature can be used to reduce the effects of such multiple testing and hence reveal weak signals in blood-based data sets. Using mice infected with lethal PR8 virus, we determined that the lung signature could be seen in blood cells of infected animals and designed a Fluidigm panel based on these data and on other information in the literature to test whether we could predict which animals would die when a group was infected at one LD50. Interestingly, while the signature gene transcripts were detected only in mice infected with PR8 and not the non-lethal H1N1 virus Tx91, and all animals showed evidence of infection from weight loss, we could not discriminate among PR8 infected animals in terms of which would die. This suggest two possibilities. Slight differences in the actual numbers of infectious particles given each animal dictate whether they survive or not in a manner that early blood signatures cannot reveal, or more interestingly, the stochastic variation among even inbred mice in their adaptive immune repertoires might dictate who suppresses potentially lethal innate inflammation before terminal pulmonary compromise. We are in the midst of further studies to distinguish between these models, with potentially important implications for understanding why some humans die from potentially lethal infections and other survive, as well as suggesting possible interventions that can tip the balance. A second project involves use of the emerging tools of systems biology to investigate the unexplored roles of many NLRs. In the course of such study, Dr. Subramanian (now heading up her own laboratory at the Institute for Systems Biology) has observed profound effects of very small changes in intracellular protein concentration on signaling through the NOD1 pathway. Under normal conditions, several miRNAs contribute to maintaining expression of NOD1 below the level leading to ligand-independent gene activation. Alternation is expression of these miRNAs is linked to an increases severity of gastric cancer, which was previously linked to NOD1. These data may be of importance in understanding how small eQTLs linked to inflammatory and autoimmune diseases operate to cause pathology. Based on these findings, we are exploring in various experimental systems whether small (1-2 fold changes in gene expression, as seen with many eQTLs), can lead to disease by imbalancing activation and negative regulatory pathways. We suspect this type of dysregulation might contribute to various autoimmune states. As part of the larger LSB group effort to better understand TLR signaling in macrophages, we have conducted fine grained time and dose studies at the single cells level and on bulk populations looking at a diverse set of downstream signaling events and effector responses. Among several intriguing observations, most striking is the discovery that contrary to many models with non-hematopoietic cells, we see a graded response to NFkB in the presence of the expected digital responses among MAPK pathways. The MAPK responses require higher concentration of TLR ligand than the NFkB responses and in the dose region in which NFkB is activated but MAPK responses are not, cells show activation of a subset of NFkB-responsive genes but also not only limited transcription and especially translation of inflammatory cytokine genes and gene products. Only when both NFkB and MAPK signals are present are inflammatory cytokines and chemokines generated. These data suggest that the macrophage response system has evolved to limit potentially damaging inflammation in the face of minor pulses of PAMP or DAMP signals engaging TLR (the steady-state), but to prepare for anti-pathogen responses under such conditions incase these weak signals are not from commensals or normal tissue turnover but from an incipient infection; if the stimulus continued to increase, as would typically be the case mainly when there is an active pathogenic infection, then the digital nature of the MAPK pathways in the context of priming through NFkB at lower ligand levels gives an immediate robust response. These data also explain observations of M. Karin that chronic NFkB activation does not lead to intestinal cancer development unless a MAPK stimulus is added. Remarkably, both humans and various mouse strains all share nearly identical dose-responses and among random human donors from the Blood Bank, the responses are nearly identical, a very unusual result in comparison to data on normal volunteers in the CHI influenza and other studies. The data from this project are also serving to drive development of new Simmune-based signaling models for the TLR pathway that has already revealed an unexpected feedback loop from p38 to Erk that is central to the dose-control we observe. This set of observation is being extended to see if the GWAS hits involving DUSPs that are key to the Simmune model help explain the basis for human inflammatory disease as a disturbance in this tightly regulated pathway. We are also proceeding with related studies using combinations of TLR inputs to develop an understanding of how macrophages decode the multiple PAMP signals of pathogens. We have also collaborated with others in the LSB in the design of a new software package (CARD) of value in prioritizing hits in large scale siRNA screens.